Third Party Funds Group - Sub project
Acronym: Hyb_Mod_Net
Start date : 01.09.2025
End date : 31.08.2028
The aim of the research project is to develop hybrid network models for district heating networks that are based on both classical differential equations and AI algorithms. These models are to be used for optimal control. Scientifically, graph databases for district heating networks are being investigated, hybrid model structures optimized and security aspects for KRITIS applications researched. Technically, the focus is on the integration of heterogeneous network data, the development of adaptive models and intelligent, energy-optimized network control. The algorithms developed will be implemented and tested in the Dortmund district heating network. As part of this sub-project, the LME is developing the necessary concepts for data integration and hybrid modeling of district heating networks. For this purpose, data such as temperature, flow rate, pressure, energy feed-in, consumption and environmental influences, which have so far been recorded in different systems (e.g. GIS, network control systems and billing systems), are brought together centrally. A graph database makes it possible to link this heterogeneous data as nodes and edges - similar to social networks - and thus create a realistic model of the grid structure. This procedure is supplemented by a hybrid system approach in which classical differential equations are supported by physics-induced AI components, such as neural networks. This approach reduces computationally intensive numerical integrations and enables rapid adaptation to new measured values, as the AI models remain small and adaptable. At the same time, the basic physical structure of the model is retained, which improves traceability.